Alibaba's Qwen Open-Sources Qwen3.8-2.4T-A95B Model
xiaosun86 · x · 2026-08-13
Alibaba's Qwen team has open-sourced the new Qwen3.8-2.4T-A95B model. The SGLang team shared empirical results showing the model's strong capability in handling complex engineering tasks; it autonomously ran for 4.5 hours to optimize its own SGLang serving with verified performance improvements. RadixArk's comparison tests indicated the model achieves 128 tok/s per user on SGLang.
Related event: Alibaba Open-Sources 2.4T Parameter Flagship Qwen3.8-Max(17 posts)→
More from Infra
- AMD Earnings Analysis & Future of Memory Summit Trends — BenBajarin · 2026-08-13
- Dev Complains About Broken GPU Rentals: Finding Instances Already in Use — abacaj · 2026-08-13
- Glean Claims Its Agent Costs 4x Less Per Task Than Claude Cowork — Scobleizer · 2026-08-13
- Enthusiasts Discuss Running Massive Qwen3.8-2.4T Models Locally — segmond · 2026-08-13
- Running DeepSeek V4 Flash Locally on 2x DGX Sparks Delivers Prosumer-Grade Performance — andrewchen · 2026-08-13
- Is Local Generative AI Worth It Anymore? Developers Struggle Against Closed Cloud Models — ImaginaryEffective63 · 2026-08-13